dsh-plugins-plan-usage

Running several coding plans side by side usually means bouncing between vendor dashboards just to check where each quota stands. dsh-plugins-plan-usage brings that visibility straight into the DeepSeek Harness web chat: a multi-line capsule badge in the bottom-right corner shows, for every active plan, the percentage consumed in the 5-hour rolling window, the weekly limit, and the monthly cap, auto-refreshing every 60 seconds with no extra tabs or doc lookups. Four plan integrations ship out of the box—OpenCode Go, GLM Z.AI (international), GLM Zhipu (domestic), and Kimi Code—each with independent on/off toggles, its own four-level credential fallback chain, and write-only key handling so no secret is ever echoed back to the browser.

The plugin is pure JavaScript with zero build step. The host side drives routes and data fetching through a plans/ registry, and the browser side drives rendering through static data tables, so neither half contains per-plan if-branches. Adding a new plan means dropping one module file into plans/ and registering it; configuration schema, settings card, badge polling, and rendering all light up automatically. The settings card matches the look and feel of Harness built-in cards and requires no modifications to Harness source code or workspace build allowlists.

Ideal for developers who subscribe to multiple coding plans and switch models frequently inside Harness—especially when a rolling window is nearing saturation and could trigger rate limits, or when shared device consumption makes monthly quota hard to predict at a glance.